Include MLE estimation of tau_0 and sigma_2

[lionelvoirol]

• Added fcts ar1_sim_geneva(), n_ll_unevenly_spaced_ar1() and ar1fit_m()l in file tsmodel.py
• Created 2 tests to ensure MLE estimation works fine with evenly and unevenly spaced AR1.
• Added time_increments() to tsmodel.py, together with tests
• eventually, the surrogate method ar1sim will point to ar1_sim_geneva(), and the current one will be relabeled ar1old for benchmarking purposes.
• For now functionalities are unaccessible from the user API ; waiting on General surrogates #228 to be completed (in stages)

[CommonClimate]

I fixed import issues in tsmodel and renamed functions as appropriate.

One remaining difference is that ar1_sim_geneva is not vectorized (it generates one version of y at a time, not an (n x p) array as ar1_sim does. For MC applications, it’s usually more efficient to vectorize, so I request that capability.

It should be easy for you to enable and test (e.g. the averages of estimated tau and sigma2 over the p realizations should be even closer to the true values).

[CommonClimate]

Thanks for these changes Lionel! I decided to pull out the part generating time increments to a dedicated function, as we might use it for something else.